Abstract
Due to the recent improvements of automatic measurement procedures in photogrammetry, multi-view 3D reconstruction technologies are becoming a favourite survey tool. Rapidly widening structure-from-motion (SfM) software packages offer significantly easier image processing workflows than traditional photogrammetry packages. However, while most orientation and surface reconstruction strategies will almost always succeed in any given task, estimating the quality of the result is, to some extent, still an open issue. An assessment of the precision and reliability of block orientation is necessary and should be included in every processing pipeline. Such a need was clearly felt from the results of close-range photogrammetric surveys of in situ full-scale and laboratory-scale experiments. In order to study the impact of the block control and the camera network design on the block orientation accuracy, a series of Monte Carlo simulations was performed. Two image block configurations were investigated: a single pseudo-normal strip and a circular highly-convergent block. The influence of surveying and data processing choices, such as the number and accuracy of the ground control points, autofocus and camera calibration was investigated. The research highlights the most significant aspects and processes to be taken into account for adequate in situ and laboratory surveys, when modern SfM software packages are used, and evaluates their effect on the quality of the results of the surface reconstruction.
Highlights
Computer vision (CV) has significantly contributed to the improvement of image-based object reconstruction, most notably to sensor calibration and orientation [1,2]
A decision must be made on whether to maintain the same principal distances (PDs) for each image and accepting that not all images are perfectly focused, or to focus each time, accepting that the PD of the camera is not the same for the whole block. The latter is chosen, because blurred images make the homologous point identification harder and less precise. Another decision must be taken for the bundle block adjustment (BBA): if it is assumed that the PD variation is negligible, a sort of mean interior orientation (IO) parameter set will be estimated; otherwise, a different focal length should be estimated for every image using a self-calibration approach
The results indicate that the precision of the ground control points (GCP) should be at least of the same order of magnitude of the ground sampling distance (GSD) to ensure that no systematic deformation and/or error propagation on the tie point coordinates arise
Summary
Computer vision (CV) has significantly contributed to the improvement of image-based object reconstruction, most notably to sensor calibration and orientation [1,2]. The high automation level of SfM software packages can hide important aspects, such as the influence of camera autofocus and the consequences of an inaccurate pre- or self-calibration procedure of the optical system [14] It seems more reasonable and more appropriate than in standard photogrammetric tasks to check the restitution error for object coordinates, though it involves the image measurement accuracy, the IO accuracy and the accuracy of check points or ground control points (GCP). Two virtual objects and camera networks are introduced representing the two real scenarios In these scenarios, the influence of various parameters on the accuracy and quality of the photogrammetric block is studied.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
